White-label AI API — ship AI in your own brand

Three things make a true white-label AI product: (1) an API you can rebrand end-to-end, (2) a dashboard you can host on your own domain with your own logo, and (3) client apps — IDE plugins, desktop apps, terminal CLIs — that look and feel like yours. Plugsky ships all three.

What is white-label AI?

A white-label AI API is a hosted inference product you can present as your own. The customer never sees the underlying provider — they see your logo, your domain, your docs, your pricing, and your support. You pay the provider, mark up the cost, and keep the margin.

Plugsky is designed for this. Everything user-facing is brand-neutral:

• API responses — the model field returns your slug (plugsky-pro), not the upstream vendor's id.
• Dashboard — a single-page static site you can fork and rebrand. Logo, colors, copy, and domain are all yours.
• Docs — MIT-licensed, written to be forked, with no vendor mentions.
• Client apps — connect any OpenAI-compatible IDE, desktop app, or CLI to your branded endpoint. No vendor lock-in.

Rebrand the API surface

The fastest rebrand is at the model id level. Every model in the ladder has a slug you control. Ship a yourbrand-pro, yourbrand-fast, yourbrand-embed and present them as your own product line.

# Your customers see this
curl -X POST https://api.yourbrand.com/v1/chat/completions \
  -H "Authorization: Bearer sk-live-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "yourbrand-pro",
    "messages": [{"role":"user","content":"Hello!"}]
  }'

# The response returns your slug, not the upstream
{
  "id": "chatcmpl-...",
  "model": "yourbrand-pro",
  "object": "chat.completion",
  "choices": [{"message":{"role":"assistant","content":"Hi!"}}]
}

OpenAI's SDKs (Python, Node, Go, .NET, Java, Ruby, Rust, PHP) all work against the same /v1 OpenAPI spec — change base_url, change the model name, and ship.

No vendor lock-in

Because Plugsky is 100% OpenAI-compatible, you can swap providers without touching customer code. The day Plugsky raises prices or misses an SLA, point base_url at OpenAI, Anthropic, or your own self-hosted model and keep going.

Customize the dashboard

The Plugsky dashboard (playground.html, dashboard.html, docs.html) is a single static SPA that talks to your branded API. Every visible string, every color, and every logo is in /articles/article.css + a top-level <nav> you ship in the HTML.

What you can rebrand in a single config file:

• Logo, wordmark, favicon
• Primary, accent, and background colors (CSS custom properties)
• Brand name in the title bar and footer
• Support email, privacy URL, terms URL
• Custom landing-page hero copy and CTAs

Host it on app.yourbrand.com with a CNAME to your Plugsky bucket, or self-host the static files in your own S3 / Cloudflare R2 / GitHub Pages.

Fork the open-source CLIs

Every terminal AI agent Plugsky integrates with is open source. The four primary recommendations are:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "yourbrand": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "YourBrand",
      "options": {
        "baseURL": "https://api.yourbrand.com/v1",
        "apiKey": "{env:YOURBRAND_API_KEY}"
      },
      "models": {
        "yourbrand-pro": { "name": "YourBrand Pro" },
        "yourbrand-max": { "name": "YourBrand Max" }
      }
    }
  }
}

export OPENAI_API_BASE=https://api.yourbrand.com/v1
export OPENAI_API_KEY=sk-live-YOUR_KEY
aider --model openai/yourbrand-pro

goose configure
→ Configure Providers
→ choose "OpenAI Compatible"
→ OPENAI_API_KEY   = sk-live-YOUR_KEY
→ OPENAI_HOST      = https://api.yourbrand.com
→ OPENAI_BASE_PATH = v1/chat/completions
→ model            = yourbrand-pro

{
  "providers": {
    "yourbrand": {
      "type": "openai",
      "base_url": "https://api.yourbrand.com/v1",
      "api_key": "$YOURBRAND_API_KEY",
      "models": [
        { "id": "yourbrand-pro", "name": "YourBrand Pro" }
      ]
    }
  }
}

Rebrand IDEs and desktop apps

Every IDE and desktop client Plugsky integrates with is configured through an OpenAI-compatible provider URL. You don't fork the IDE — you ship a config preset the user pastes in.

Visual Studio Code (Continue, Cline, Roo Code)

Each extension reads a JSON or YAML config. Ship a one-line snippet:

{
  "models": [{
    "title": "YourBrand Pro",
    "provider": "openai",
    "model": "yourbrand-pro",
    "apiBase": "https://api.yourbrand.com/v1",
    "apiKey": "sk-live-..."
  }]
}

JetBrains AI Assistant

Settings → Tools → AI Assistant → Custom LLM provider → OpenAI-compatible → point at your endpoint.

Desktop apps (Chatbox, Jan, LM Studio, Msty, Open WebUI, LibreChat, Raycast)

All seven ship a Custom OpenAI Endpoint field. Plug in your base URL + key + model id; the app routes every call to your brand.

Model licensing — what can you rebrand?

Open-weight models (Llama, Mistral, Qwen, DeepSeek, Gemma) ship under permissive licenses that allow commercial rebranding. Closed models (the ones Plugsky proxies from NVIDIA) carry their own terms — check the upstream license before shipping a white-label.

For full control, host your own open-weights on your own GPU. Plugsky's inference layer accepts any OpenAI-compatible server — point it at vLLM, llama.cpp, or TGI running in your VPC.

Frequently asked questions